Knee osteoarthritis (OA) is a major cause of pain, functional limitation and disability and among the most costly musculoskeletal conditions. A burgeoning population with knee OA and poor clinical outcomes in the absence of effective treatments are key drivers of the soaring rates and costs of knee replacement. Knee OA pathology and clinical outcomes typically unfold over decades and have a highly varied time course. It is a priority to understand the range of factors that contribute to poor and good outcomes in knee OA, but there are critical gaps in knowledge about the long-term course of the disease and its determinants. It is sensible to target prevention and treatment on those most at risk for poor outcomes. However, there is a paucity of longitudinal studies long enough (?10 years) with frequent assessment using standardized measures to capture the full trajectory, range and variability of outcomes and investigate their determinants. In addition, investigation of these questions is hampered by baseline heterogeneity in knee OA severity and impact, which can be a source of imprecision and bias in observational studies of disease prognosis. Started in 2002, the OAI is a unique cohort study of 4796 persons with or at risk for knee OA that has uniform, rich clinical and imaging data from annual assessments and has been followed comprehensively for up to 8 years with good retention. We now propose to continue assessment of outcomes of knee OA for up to 15 years after baseline, primarily by phone and mail, an efficient and less burdensome approach that will yield more complete follow-up than clinic visits alone in this aging cohort. A brief clinic visit in a subset of participants will enale us to study performance measures of function as a long-term outcome. Our goal is to take advantage of this unprecedented opportunity for long-term follow-up to describe the full trajectory and probability of outcomes at different stages of disease, to identify vulnerable and protected phenotypes in its long-term course and investigate potentially modifiable predictors of these outcomes. Outcomes will span a range of health domains, including: knee-OA related (pain, functional limitation and performance, OA global impact, knee replacement), disability/participation and general health. The extended follow-up provides the opportunity for an analysis design that reduces the risk of common sources of bias and imprecision in studies of disease prognosis by allowing the use of changes from baseline to year 4 to define inception events, such as the transition to more advanced stages of knee pain or structural OA or the development of favorable and unfavorable trajectories of risk factors (knee OA-related, physical performance, general health and psychosocial variables). These changes will then be evaluated for their association with subsequent outcomes occurring from year 5 to year 15 in analyses conditioned on a range of baseline knee OA-related and other covariates with a minimal risk of collider stratification bias. Using this approach we will investigate questions that can inform emerging strategies for prevention and treatment that are focused on improving outcomes in persons with knee OA. These include describing the probabilities of both poor and good long-term outcomes at different stages of knee pain and structural OA, comparing outcomes in those with recent transitions to more advanced stages with those whose course has been stable, and identifying the modifiable determinants of long-term outcomes at different stages of disease and whether these differ by stage, thus suggesting the need for tailored interventions to improve outcomes.